(usagi-users 02174) ping6: recvmsg: No route to host

["Philip Ross" <psr@xxxxxxxxxxxxxxxxxxxx>]

I am using a Redhat 8.0 box (2.4.18-19.8.0 kernel, non-usagi) to experiment
with IPv6.
I have a /64 block allocated from uk6x.com using an IPv6-in-IPv4 (sit)
tunnel.

Everything is working fine apart from pings and traceroutes from the linux
box.  I can make tcp connections and send/receive udp packets fine.  Other
local machines have no problems with pings or traceroutes routing through
the linux box.  I can ping the linux box fine from an external host.

When pinging a remote host from the linux box I get the following results:

# ping6 www.uk6x.com
PING www.uk6x.com(kesey.uk6x.com) 56 data bytes
ping: recvmsg: No route to host
ping: recvmsg: No route to host
ping: recvmsg: No route to host

--- www.uk6x.com ping statistics ---
3 packets transmitted, 0 received, 100% loss, time 2003ms


If I reduce the packet size to 25 bytes, the pings work fine:

# ping6 -s 17 www.uk6x.com
PING www.uk6x.com(kesey.uk6x.com) 17 data bytes
25 bytes from kesey.uk6x.com: icmp_seq=1 ttl=62 time=71.1 ms
25 bytes from kesey.uk6x.com: icmp_seq=2 ttl=62 time=28.2 ms
25 bytes from kesey.uk6x.com: icmp_seq=3 ttl=62 time=27.4 ms

--- www.uk6x.com ping statistics ---
3 packets transmitted, 3 received, 0% loss, time 2004ms
rtt min/avg/max/mdev = 27.483/42.297/71.144/20.400 ms


Doing a tcpdump on sit1 shows that the 56 byte pings are being replied to:

# tcpdump -i sit1
tcpdump: WARNING: sit1: no IPv4 address assigned
tcpdump: listening on sit1
22:25:10.022171 2001:618:400::xxxx:yyyy > kesey.uk6x.com: icmp6: echo
request
22:25:10.056299 kesey.uk6x.com > 2001:618:400::xxxx:yyyy: icmp6: echo reply
22:25:11.024222 2001:618:400::xxxx:yyyy > kesey.uk6x.com: icmp6: echo
request
22:25:11.056286 kesey.uk6x.com > 2001:618:400::xxxx:yyyy: icmp6: echo reply
22:25:12.025995 2001:618:400::xxxx:yyyy > kesey.uk6x.com: icmp6: echo
request
22:25:12.059296 kesey.uk6x.com > 2001:618:400::xxxx:yyyy: icmp6: echo reply


If I ping a local machine from the linux box using the global IPv6 address I
get the same problem.  If I use the local machine's link local address
everything works fine.

My routing table:

# route -A inet6
Kernel IPv6 routing table
Destination                                 Next Hop
Flags Metric Ref    Use Iface
::1/128                                     ::
U     0      1        0 lo
::127.0.0.1/128                             ::
U     0      0        0 lo
::192.168.1.100/128                         ::
U     0      0        0 lo
::/96                                       ::
U     256    0        0 sit0
2001:618:400::xxxx:yyyy/128                 ::
U     0      35       0 lo
2001:618:400:zzzz::/128                     ::
U     0      18       0 lo
2001:618:400:zzzz::100/128                  ::
U     0      2        0 lo
2001:618:400:zzzz::/64                      ::
UA    256    0        0 eth0
2000::/3                                    fe80::c171:3a50
UG    1      65       0 sit1
fe80::c0a8:164/128                          ::
U     0      69       0 lo
fe80::250:56ff:fe40:5bb5/128                fe80::250:56ff:fe40:5bb5
UAC   0      1        1 eth0
fe80::2a0:ccff:fed0:fb9/128                 ::
U     0      33       0 lo
fe80::/10                                   ::
UA    256    0        0 eth0
fe80::/10                                   ::
UA    256    102       0 sit1
ff02::1/128                                 ff02::1
UAC   0      6        1 eth0
ff00::/8                                    ::
UA    256    0        0 eth0
ff00::/8                                    ::
UA    256    0        0 sit1
::/0                                        fe80::c171:3a50
UG    1      0        0 sit1
::/0                                        ::
UA    256    0        0 eth0



Has anyone got any ideas as to what might be going on here? Are there any
known incompatibilities with particular ethernet cards (I'm using a Netgear
FA3?? card - Lite-On Communications Inc LNE100TX (rev 32) - tulip kernel
module)?

Thanks,

Phil